#115 Aspell crashes on some Uzbek words

[Mashrab Kuvatov]

I have Aspell-060.1.1 and Uzbek dictionary
aspell6-uz-0.02-1. It crashes on some Uzbek words with
the following error message:

aspell: modules/speller/default/suggest.cpp:519: void
<unnamed>::Working::add_nearmiss(__gnu_cxx::__normal_iterator<const
aspeller::Dict* const*, std::vector<const aspeller::Dict*,
std::allocator<const aspeller::Dict*> > >, const
aspeller::WordEntry&, const char*, int, int, bool): Assertion
`w.word_size == strlen(w.word)' failed.
Aborted

The test cases are attached (Unicode encoded files). To
reproduce run aspell -d uz -c <file-name>.

[Mashrab Kuvatov]

Test case for Aspell (Unicode encoded file)

[Kevin Atkinson]

Logged In: YES
user_id=6591

You didn't attach anything. Probably forgot to check the
checkbox after " Check to Upload and Attach a File".

[Mashrab Kuvatov]

Logged In: YES
user_id=902405

I wanted to put those Uzbek words into separate files, but SF complained
that the files were too small. So, I put them into one file.

[Kevin Atkinson]

Logged In: YES
user_id=6591

I can't seam to reproduce the problem.

You may be using a dictionary that was compiled with a
version of Aspell before 0.60. Try recompiling the
dictionary and make sure you are using it.

[Mashrab Kuvatov]

[Mashrab Kuvatov]

Logged In: YES
user_id=902405

I think I know why you cannot reproduce the problem. You have
to have .aspell.uz.prepl in $HOME. I tried to remove it and then
spell-check, Aspell did not crash. I attached my .aspell.uz.prepl
to this bug report.

Thanks.

[Kevin Atkinson]

Logged In: YES
user_id=6591

OK confirmed. I will look into it.

[Kevin Atkinson]

Logged In: YES
user_id=6591

Fixed. Patch attached.

[Kevin Atkinson]

[Mashrab Kuvatov]

Logged In: YES
user_id=902405

Kevin, I tried the patch. It works. Thanks.

Just wondering. the attached .aspell.uz.prepl file has three lines,
but the first line is "personal_repl-1.1 uz 0 utf-8". Shouldn't it be
"personal_repl-1.1 uz 3 utf-8"?

[Kevin Atkinson]

Logged In: YES
user_id=6591

> Shouldn't it be "personal_repl-1.1 uz 3 utf-8"

The count is currently unused. Even if I do use it in a
future version it will only act as a hint and thus doesn't
have to be accurate.